[1.6.5] - 2026-08-21 Changed get_el_amt_dict and get_components_by_element now look up each solute's oxi_state_guesses once per species instead of redundantly re-fetching it on every element of that species. Output is unchanged; the two methods run roughly 8% and 11% faster, respectively, on a 47-species seawater solution. (@rkingsbury) CHANGED BEHAVIOR Solution.__mul__ / Solution.__truediv__: scalar sol * factor and sol / factor no longer mutate the original Solution and return an alias of it. They now return a new, independently scaled Solution, leaving the original unchanged (consistent with Python numeric-type semantics). In-place *= / /= retain their documented mutate-in-place behavior (now via explicit __imul__ / __itruediv__), and factor * sol is now supported via __rmul__. (#461, @youdie006) Fixed Solution.alkalinity is now invariant to how the equilibrium engine speciates the solution. The conservative-ion contributions are computed from the total analytical concentration of each constituent (via get_total_amount) rather than from free-ion concentrations. Previously, engines such as PHREEQC that form ion pairs (e.g. NaSO4[-1], MgSO4(aq), CaSO4(aq)) would reduce the free SO4[-2] concentration during equilibrate(), spuriously inflating the alkalinity of seawater from ~117 to ~753 mg/L as CaCO3. Credit to @jjstickel for reporting. (#458, @rkingsbury) .gitignore now correctly excludes ipynb_checkpoint files. (@rkingsbury; @jjstickel)